Pub Date : 2025-10-11DOI: 10.1016/j.aiia.2025.10.005
Houkang Jiang , Jizhan Liu , Xiaojie Lei , Baocheng Xu , Yucheng Jin
In unstructured orchard environments, factors such as complex lighting, fruit occlusion, and fruit clustering significantly reduce the accuracy of apple detection and 3D localization in robotic harvesting systems. To enhance the perception and reconstruction of occluded fruits, this paper proposes a multi-stage fusion framework for high-precision and robust processing, from image enhancement to 3D reconstruction. First, an adaptive image enhancement algorithm based on the HSV color space is employed to effectively alleviate image degradation caused by uneven lighting. Then, an improved Mask R-CNN with a dual-attention mechanism (SE-CBAM) is introduced to achieve scale-adaptive fruit segmentation under occlusion conditions. Next, a hierarchical point cloud purification strategy combining depth clustering and geometric feature analysis is applied to remove branch and leaf interference. Finally, a two-step RANSAC-LM spherical fitting algorithm is used to quickly and accurately recover the 3D shape of the fruit from incomplete point clouds. Experimental results show that the proposed method achieves Mask IoUs of 0.89, 0.82, and 0.65 under mild, moderate, and severe occlusion, respectively, with the lowest 3D localization error of 0.42 cm and an overall processing frame rate of 20 FPS. In real orchard environments, the harvesting success rate under occlusion conditions reaches up to 82.7 %, significantly outperforming traditional point cloud centroid and 2D positioning methods. This study provides an efficient, robust, and real-time deployable visual solution for fruit localization and robotic harvesting in complex orchard environments.
{"title":"Multi-stage fusion of dual attention mask R-CNN and geometric filtering for fast and accurate localization of occluded apples","authors":"Houkang Jiang , Jizhan Liu , Xiaojie Lei , Baocheng Xu , Yucheng Jin","doi":"10.1016/j.aiia.2025.10.005","DOIUrl":"10.1016/j.aiia.2025.10.005","url":null,"abstract":"<div><div>In unstructured orchard environments, factors such as complex lighting, fruit occlusion, and fruit clustering significantly reduce the accuracy of apple detection and 3D localization in robotic harvesting systems. To enhance the perception and reconstruction of occluded fruits, this paper proposes a multi-stage fusion framework for high-precision and robust processing, from image enhancement to 3D reconstruction. First, an adaptive image enhancement algorithm based on the HSV color space is employed to effectively alleviate image degradation caused by uneven lighting. Then, an improved Mask R-CNN with a dual-attention mechanism (SE-CBAM) is introduced to achieve scale-adaptive fruit segmentation under occlusion conditions. Next, a hierarchical point cloud purification strategy combining depth clustering and geometric feature analysis is applied to remove branch and leaf interference. Finally, a two-step RANSAC-LM spherical fitting algorithm is used to quickly and accurately recover the 3D shape of the fruit from incomplete point clouds. Experimental results show that the proposed method achieves Mask IoUs of 0.89, 0.82, and 0.65 under mild, moderate, and severe occlusion, respectively, with the lowest 3D localization error of 0.42 cm and an overall processing frame rate of 20 FPS. In real orchard environments, the harvesting success rate under occlusion conditions reaches up to 82.7 %, significantly outperforming traditional point cloud centroid and 2D positioning methods. This study provides an efficient, robust, and real-time deployable visual solution for fruit localization and robotic harvesting in complex orchard environments.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 1","pages":"Pages 187-205"},"PeriodicalIF":12.4,"publicationDate":"2025-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145361969","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-03DOI: 10.1016/j.aiia.2025.10.001
Zhijian Chen , Jianjun Yin , Sheikh Muhammad Farhan , Lu Liu , Ding Zhang , Maile Zhou , Junhui Cheng
As automation becomes increasingly adopted to mitigate labor shortages and boost productivity, autonomous technologies such as tractors, drones, and robotic devices are being utilized for various tasks that include plowing, seeding, irrigation, fertilization, and harvesting. Successfully navigating these changing agricultural landscapes necessitates advanced sensing, control, and navigation systems that can adapt in real time to guarantee effective and safe operations. This review focuses on obstacle avoidance systems in autonomous farming machinery, highlighting multi-functional capabilities within intricate field settings. It analyzes various sensing technologies, LiDAR, visual cameras, radar, ultrasonic sensors, GPS/GNSS, and inertial measurement units (IMU) for their individual and collective contributions to precise obstacle detection in fluctuating field conditions. The review examines the potential of multi-sensor fusion to enhance detection accuracy and reliability, with a particular emphasizing on achieving seamless obstacle recognition and response. It addresses recent advancements in control and navigation systems, particularly focusing on path-planning algorithms and real-time decision-making. It enables autonomous systems to adjust dynamically across multi-functional agricultural environments. The methodologies used for path planning, including adaptive and learning-based strategies, are discussed for their ability to optimize navigation in complicated field conditions. Real-time decision-making frameworks are similarly evaluated for their capacity to provide prompt, data-driven reactions to changing obstacles, which is critical for maintaining operational efficiency. Moreover, this review discusses environmental and topographical challenges like variable terrain, unpredictable weather, complex crop arrangements, and interference from co-located machinery that hinder obstacle detection and necessitate adaptive, resilient system responses. In addition, the paper emphasizes future research opportunities, highlighting the significance of advancements in multi-sensor fusion, deep learning for perception, adaptive path planning, model-free control strategies, artificial intelligence, and energy-efficient designs. Enhancing obstacle avoidance systems enables autonomous agricultural machinery to transform modern farming by increasing efficiency, precision, and sustainability. The review highlights the potential of these technologies to support global efforts for sustainable agriculture and food security, aligning agricultural innovation with the needs of a swiftly growing population.
{"title":"A comprehensive review of obstacle avoidance for autonomous agricultural machinery in multi-operational environment","authors":"Zhijian Chen , Jianjun Yin , Sheikh Muhammad Farhan , Lu Liu , Ding Zhang , Maile Zhou , Junhui Cheng","doi":"10.1016/j.aiia.2025.10.001","DOIUrl":"10.1016/j.aiia.2025.10.001","url":null,"abstract":"<div><div>As automation becomes increasingly adopted to mitigate labor shortages and boost productivity, autonomous technologies such as tractors, drones, and robotic devices are being utilized for various tasks that include plowing, seeding, irrigation, fertilization, and harvesting. Successfully navigating these changing agricultural landscapes necessitates advanced sensing, control, and navigation systems that can adapt in real time to guarantee effective and safe operations. This review focuses on obstacle avoidance systems in autonomous farming machinery, highlighting multi-functional capabilities within intricate field settings. It analyzes various sensing technologies, LiDAR, visual cameras, radar, ultrasonic sensors, GPS/GNSS, and inertial measurement units (IMU) for their individual and collective contributions to precise obstacle detection in fluctuating field conditions. The review examines the potential of multi-sensor fusion to enhance detection accuracy and reliability, with a particular emphasizing on achieving seamless obstacle recognition and response. It addresses recent advancements in control and navigation systems, particularly focusing on path-planning algorithms and real-time decision-making. It enables autonomous systems to adjust dynamically across multi-functional agricultural environments. The methodologies used for path planning, including adaptive and learning-based strategies, are discussed for their ability to optimize navigation in complicated field conditions. Real-time decision-making frameworks are similarly evaluated for their capacity to provide prompt, data-driven reactions to changing obstacles, which is critical for maintaining operational efficiency. Moreover, this review discusses environmental and topographical challenges like variable terrain, unpredictable weather, complex crop arrangements, and interference from co-located machinery that hinder obstacle detection and necessitate adaptive, resilient system responses. In addition, the paper emphasizes future research opportunities, highlighting the significance of advancements in multi-sensor fusion, deep learning for perception, adaptive path planning, model-free control strategies, artificial intelligence, and energy-efficient designs. Enhancing obstacle avoidance systems enables autonomous agricultural machinery to transform modern farming by increasing efficiency, precision, and sustainability. The review highlights the potential of these technologies to support global efforts for sustainable agriculture and food security, aligning agricultural innovation with the needs of a swiftly growing population.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 1","pages":"Pages 139-163"},"PeriodicalIF":12.4,"publicationDate":"2025-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145264762","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-02DOI: 10.1016/j.aiia.2025.10.003
Liuyan Feng , Changsu Xu , Han Tang , Zhongcai Wei , Xiaodong Guan , Jingcheng Xu , Mingjin Yang , Yunwu Li
With the rapid advancement of information technology, the intelligent and unmanned applications of agricultural machinery and equipment have become a central focus of current research. Navigation technology is central to achieving autonomous driving in agricultural machinery and plays a key role in advancing intelligent agriculture. However, although some studies have reviewed aspects of agricultural machinery navigation technologies, a comprehensive and systematic overview that clearly outlines the developmental trajectory of these technologies is still lacking. At the same time, there is an urgent need to break through traditional navigation frameworks to address the challenges posed by complex agricultural environments. Addressing this gap, this study provides a comprehensive overview of the evolution of navigation technologies in agricultural machinery, categorizing them into three stages: assisted navigation, autonomous navigation, and intelligent navigation, based on the level of autonomy in agricultural machinery. Special emphasis is placed on the brain-inspired navigation technology, which is an important branch of intelligent navigation and has attracted widespread attention as an emerging direction. It innovatively mimics the cognitive and learning abilities of the brain, demonstrating high adaptability and robustness to better handle uncertainty and complex environments. Importantly, this paper innovatively explores six potential applications of brain-inspired navigation technology in the agricultural field, highlighting its significant potential to enhance the intelligence of agricultural machinery. The review concludes by discussing current limitations and future research directions. The findings of this study aim to guide the development of more intelligent, adaptive, and resilient navigation systems, accelerating the transformation toward fully autonomous agricultural operations.
{"title":"Application of navigation technology in agricultural machinery: A review and prospects","authors":"Liuyan Feng , Changsu Xu , Han Tang , Zhongcai Wei , Xiaodong Guan , Jingcheng Xu , Mingjin Yang , Yunwu Li","doi":"10.1016/j.aiia.2025.10.003","DOIUrl":"10.1016/j.aiia.2025.10.003","url":null,"abstract":"<div><div>With the rapid advancement of information technology, the intelligent and unmanned applications of agricultural machinery and equipment have become a central focus of current research. Navigation technology is central to achieving autonomous driving in agricultural machinery and plays a key role in advancing intelligent agriculture. However, although some studies have reviewed aspects of agricultural machinery navigation technologies, a comprehensive and systematic overview that clearly outlines the developmental trajectory of these technologies is still lacking. At the same time, there is an urgent need to break through traditional navigation frameworks to address the challenges posed by complex agricultural environments. Addressing this gap, this study provides a comprehensive overview of the evolution of navigation technologies in agricultural machinery, categorizing them into three stages: assisted navigation, autonomous navigation, and intelligent navigation, based on the level of autonomy in agricultural machinery. Special emphasis is placed on the brain-inspired navigation technology, which is an important branch of intelligent navigation and has attracted widespread attention as an emerging direction. It innovatively mimics the cognitive and learning abilities of the brain, demonstrating high adaptability and robustness to better handle uncertainty and complex environments. Importantly, this paper innovatively explores six potential applications of brain-inspired navigation technology in the agricultural field, highlighting its significant potential to enhance the intelligence of agricultural machinery. The review concludes by discussing current limitations and future research directions. The findings of this study aim to guide the development of more intelligent, adaptive, and resilient navigation systems, accelerating the transformation toward fully autonomous agricultural operations.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 1","pages":"Pages 94-123"},"PeriodicalIF":12.4,"publicationDate":"2025-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145264764","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-02DOI: 10.1016/j.aiia.2025.09.001
Reem Abukmeil , Ahmad Al-Mallahi , Felipe Campelo
The ability to sense nutrient status in potato plants using spectroscopy has several merits including the ability to proactively respond to deficiencies of certain elements. While research so far has focused on finding spectral signatures of elements based on their foliar reflectance, the influence of the spectral signatures of the elements on each other in estimating their concentrations in the plant has not been investigated. This work presents a pipeline of stacked regression models capable of accurately estimating nutrient concentrations based on the foliar reflectance. A data set was built from 179 samples of petioles collected across two growing seasons, consisting of the chemical concentrations of 11 nutrients with spectral reflectance values between 400 and 2500 nm. The pipeline consisted of a base layer composed of a multiple univariate linear Lasso regression models to find the initial independent signatures of each nutrient, followed by a layer of nonlinear models to correlate these signatures and account for their interdependencies before finalizing the estimation. The results show that adding this second layer improved estimation performance for 10 and 9 nutrients out of 12 in the dried and fresh mode, respectively, with large improvements in predictive performance for some critical micronutrients such as Zn, Fe, and Al.
{"title":"Multivariate stacked regression pipeline to estimate correlated macro and micronutrients in potato plants using visible and near-infrared reflectance spectra","authors":"Reem Abukmeil , Ahmad Al-Mallahi , Felipe Campelo","doi":"10.1016/j.aiia.2025.09.001","DOIUrl":"10.1016/j.aiia.2025.09.001","url":null,"abstract":"<div><div>The ability to sense nutrient status in potato plants using spectroscopy has several merits including the ability to proactively respond to deficiencies of certain elements. While research so far has focused on finding spectral signatures of elements based on their foliar reflectance, the influence of the spectral signatures of the elements on each other in estimating their concentrations in the plant has not been investigated. This work presents a pipeline of stacked regression models capable of accurately estimating nutrient concentrations based on the foliar reflectance. A data set was built from 179 samples of petioles collected across two growing seasons, consisting of the chemical concentrations of 11 nutrients with spectral reflectance values between 400 and 2500 nm. The pipeline consisted of a base layer composed of a multiple univariate linear Lasso regression models to find the initial independent signatures of each nutrient, followed by a layer of nonlinear models to correlate these signatures and account for their interdependencies before finalizing the estimation. The results show that adding this second layer improved estimation performance for 10 and 9 nutrients out of 12 in the dried and fresh mode, respectively, with large improvements in predictive performance for some critical micronutrients such as Zn, Fe, and Al.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 1","pages":"Pages 85-93"},"PeriodicalIF":12.4,"publicationDate":"2025-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145264765","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-02DOI: 10.1016/j.aiia.2025.10.002
Tiantian Jiang , Liang Li , Zhen Zhang , Xun Yu , Yanqin Zhu , Liming Li , Yadong Liu , Yali Bai , Ziqian Tang , Shuaibing Liu , Yan Zhang , Zheng Duan , Dameng Yin , Xiuliang Jin
Maize seedling count and leaf age are critical indicators of early growth status, essential for effective field management and breeding variety selection. Traditional field monitoring methods are time-consuming, labor-intensive, and prone to subjective errors. Recently, deep learning-based object detection models have gained attention in crop seedling counting. However, many of these models exhibit high computational complexity and implementation costs, making field deployment challenging. Moreover, maize leaf age monitoring in field environments is barely investigated. Therefore, this study proposes two lightweight models, YOLOv8n-Light-Pruned (YOLOv8n-LP) and YOLOv11n-Light-Pruned (YOLOv11n-LP), for monitoring maize seedling count and leaf age in field RGB images. Our proposed models are improved from YOLOv8n and YOLOv11n by incorporating the DAttention mechanism, an improved BiFPN, an EfficientHead, and layer-adaptive magnitude-based pruning. The improvement in model complexity and model efficiency was significant, with the number of parameters reduced by over 73 % and model efficiency upgraded by up to 42.9 % depending on the device computation power. High accuracy was achieved in seedling counting (YOLOv8n-LP/ YOLOv11n-LP: AP = 0.968/0.969, R2 = 0.91/0.94, rRMSE = 6.73 %/5.59 %), with significantly reduced model size (YOLOv8n-LP/ YOLOv11n-LP: parameters = 0.8 M/0.7 M, trained model size = 1.8 MB/1.7 MB). The robustness was validated across datasets with varying leaf ages (rRMSE = 4.07 % – 7.27 %), resolutions (rRMSE = 3.06 % – 6.28 %), seedling compositions (rRMSE = 1.09 % – 9.29 %), and planting densities (rRMSE = 3.38 % – 10.82 %). Finally, by integrating plant counting and leaf age estimation, the proposed models demonstrated high accuracy in leaf age detection using near-ground images (YOLOv8n-LP/ YOLOv11n-LP: rRMSE = 5.73 %/7.54 %) and UAV images (rRMSE = 9.24 %/14.44 %). The results demonstrate that the proposed models excel in detection accuracy, deployment efficiency, and adaptability to complex field environments, providing robust support for practical applications in precision agriculture.
{"title":"YOLO-light-pruned: A lightweight model for monitoring maize seedling count and leaf age using near-ground and UAV RGB images","authors":"Tiantian Jiang , Liang Li , Zhen Zhang , Xun Yu , Yanqin Zhu , Liming Li , Yadong Liu , Yali Bai , Ziqian Tang , Shuaibing Liu , Yan Zhang , Zheng Duan , Dameng Yin , Xiuliang Jin","doi":"10.1016/j.aiia.2025.10.002","DOIUrl":"10.1016/j.aiia.2025.10.002","url":null,"abstract":"<div><div>Maize seedling count and leaf age are critical indicators of early growth status, essential for effective field management and breeding variety selection. Traditional field monitoring methods are time-consuming, labor-intensive, and prone to subjective errors. Recently, deep learning-based object detection models have gained attention in crop seedling counting. However, many of these models exhibit high computational complexity and implementation costs, making field deployment challenging. Moreover, maize leaf age monitoring in field environments is barely investigated. Therefore, this study proposes two lightweight models, YOLOv8n-Light-Pruned (YOLOv8n-LP) and YOLOv11n-Light-Pruned (YOLOv11n-LP), for monitoring maize seedling count and leaf age in field RGB images. Our proposed models are improved from YOLOv8n and YOLOv11n by incorporating the DAttention mechanism, an improved BiFPN, an EfficientHead, and layer-adaptive magnitude-based pruning. The improvement in model complexity and model efficiency was significant, with the number of parameters reduced by over 73 % and model efficiency upgraded by up to 42.9 % depending on the device computation power. High accuracy was achieved in seedling counting (YOLOv8n-LP/ YOLOv11n-LP: AP = 0.968/0.969, R<sup>2</sup> = 0.91/0.94, rRMSE = 6.73 %/5.59 %), with significantly reduced model size (YOLOv8n-LP/ YOLOv11n-LP: parameters = 0.8 M/0.7 M, trained model size = 1.8 MB/1.7 MB). The robustness was validated across datasets with varying leaf ages (rRMSE = 4.07 % – 7.27 %), resolutions (rRMSE = 3.06 % – 6.28 %), seedling compositions (rRMSE = 1.09 % – 9.29 %), and planting densities (rRMSE = 3.38 % – 10.82 %). Finally, by integrating plant counting and leaf age estimation, the proposed models demonstrated high accuracy in leaf age detection using near-ground images (YOLOv8n-LP/ YOLOv11n-LP: rRMSE = 5.73 %/7.54 %) and UAV images (rRMSE = 9.24 %/14.44 %). The results demonstrate that the proposed models excel in detection accuracy, deployment efficiency, and adaptability to complex field environments, providing robust support for practical applications in precision agriculture.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 1","pages":"Pages 164-186"},"PeriodicalIF":12.4,"publicationDate":"2025-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145324222","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-02DOI: 10.1016/j.aiia.2025.10.004
Li Song , Jiliang Zhao , Yahui Li , Linru Liu , Jianzhao Duan , Li He , Yonghua Wang , Tiancai Guo , Wei Feng
Powdery mildew seriously hinders photosynthesis and nutrient accumulation in wheat, and its early detection holds the key to enhancing control efficacy. In this research, solar-induced chlorophyll fluorescence (SIF) parameters were derived from radiance and reflectance data, while vegetation indices (VI) were computed using reflectance. A suite of feature selection methods, including shadow feature (Boruta), feature selection (ReliefF), minimum redundancy maximum correlation (mRMR), and random forest (RF). Models were developed on the back propagation (BP) neural network, support vector regression (SVR), and partial least squares regression (PLSR). Furthermore, a stacking ensemble strategy was adopted, utilizing RF and decision tree (DT) algorithms as meta-models to integrate the predictions from base models. The findings revealed that the Boruta method selected a well-balanced number of feature parameters with normalized weights. The multi-source model (SIF + VI) is superior to the single-source model (SIF or VI). The BP model exhibited high accuracy in wheat disease monitoring, particularly during the initial infection phases. The multi-regressor stacked with RF ensemble model (MRSRF) generally surpassed the multi-regressor stacked with DT ensemble model (MRSDT), especially in the initial infection stage, where the MRSRF model's average R2 was 13.03 % higher than that of the BP model. To validate these conclusions, reflectance data simulated by the PROSAIL model (PROSPECT and SAIL) were utilized. The Boruta-MRSRF model demonstrated exceptional advantages in early detection, achieving an R2 greater than 0.90 at all infection stages. This study provides effective ideas and methods for the active prevention and control of crop diseases, which are of great significance for ensuring agricultural production.
{"title":"Early detection of wheat powdery mildew: A multi-source in situ remote sensing approach enabled by stacked ensemble learning","authors":"Li Song , Jiliang Zhao , Yahui Li , Linru Liu , Jianzhao Duan , Li He , Yonghua Wang , Tiancai Guo , Wei Feng","doi":"10.1016/j.aiia.2025.10.004","DOIUrl":"10.1016/j.aiia.2025.10.004","url":null,"abstract":"<div><div>Powdery mildew seriously hinders photosynthesis and nutrient accumulation in wheat, and its early detection holds the key to enhancing control efficacy. In this research, solar-induced chlorophyll fluorescence (SIF) parameters were derived from radiance and reflectance data, while vegetation indices (VI) were computed using reflectance. A suite of feature selection methods, including shadow feature (Boruta), feature selection (ReliefF), minimum redundancy maximum correlation (mRMR), and random forest (RF). Models were developed on the back propagation (BP) neural network, support vector regression (SVR), and partial least squares regression (PLSR). Furthermore, a stacking ensemble strategy was adopted, utilizing RF and decision tree (DT) algorithms as meta-models to integrate the predictions from base models. The findings revealed that the Boruta method selected a well-balanced number of feature parameters with normalized weights. The multi-source model (SIF + VI) is superior to the single-source model (SIF or VI). The BP model exhibited high accuracy in wheat disease monitoring, particularly during the initial infection phases. The multi-regressor stacked with RF ensemble model (MRSRF) generally surpassed the multi-regressor stacked with DT ensemble model (MRSDT), especially in the initial infection stage, where the MRSRF model's average R<sup>2</sup> was 13.03 % higher than that of the BP model. To validate these conclusions, reflectance data simulated by the PROSAIL model (PROSPECT and SAIL) were utilized. The Boruta-MRSRF model demonstrated exceptional advantages in early detection, achieving an R<sup>2</sup> greater than 0.90 at all infection stages. This study provides effective ideas and methods for the active prevention and control of crop diseases, which are of great significance for ensuring agricultural production.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 1","pages":"Pages 124-138"},"PeriodicalIF":12.4,"publicationDate":"2025-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145264763","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper presents a comprehensive review of recent advancements in intelligent monitoring systems within the precision viticulture sector. These systems have the potential to make agricultural production more efficient and ensure the adoption of sustainable practices to increase food production and meet growing global demand while maintaining high-quality standards. The review examines core components of non-destructive imaging-based monitoring systems in vineyards, focusing on sensors, tasks, and data processing methodologies. Particular emphasis is placed on solutions designed for practical, in-field deployment. The analysis revealed that the most commonly used sensors are RGB cameras and that the most widespread analysis focuses on grape bunches, as they provide information on both the quality and quantity of the harvest. Regarding the image processing methods, it emerged that those based on deep learning are the most adopted. In addition, a detailed analysis highlights the main technical and practical limitations in real-world scenarios, such as the management of computational resources, the need for large datasets, and the difficulties in interpreting the results. The paper concludes with an in-depth discussion of the challenges and open research questions, providing insights into potential future directions for intelligent monitoring systems in precision viticulture. These include the continued exploration of sensors to balance ease of use and accuracy, the development of generalizable methods, experimentation in real-world scenarios, and collaboration between experts for practical solutions.
{"title":"A perspective analysis of imaging-based monitoring systems in precision viticulture: Technologies, intelligent data analyses and research challenges","authors":"Annaclaudia Bono , Cataldo Guaragnella , Tiziana D'Orazio","doi":"10.1016/j.aiia.2025.08.001","DOIUrl":"10.1016/j.aiia.2025.08.001","url":null,"abstract":"<div><div>This paper presents a comprehensive review of recent advancements in intelligent monitoring systems within the precision viticulture sector. These systems have the potential to make agricultural production more efficient and ensure the adoption of sustainable practices to increase food production and meet growing global demand while maintaining high-quality standards. The review examines core components of non-destructive imaging-based monitoring systems in vineyards, focusing on sensors, tasks, and data processing methodologies. Particular emphasis is placed on solutions designed for practical, in-field deployment. The analysis revealed that the most commonly used sensors are RGB cameras and that the most widespread analysis focuses on grape bunches, as they provide information on both the quality and quantity of the harvest. Regarding the image processing methods, it emerged that those based on deep learning are the most adopted. In addition, a detailed analysis highlights the main technical and practical limitations in real-world scenarios, such as the management of computational resources, the need for large datasets, and the difficulties in interpreting the results. The paper concludes with an in-depth discussion of the challenges and open research questions, providing insights into potential future directions for intelligent monitoring systems in precision viticulture. These include the continued exploration of sensors to balance ease of use and accuracy, the development of generalizable methods, experimentation in real-world scenarios, and collaboration between experts for practical solutions.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 1","pages":"Pages 62-84"},"PeriodicalIF":12.4,"publicationDate":"2025-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145048407","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-21DOI: 10.1016/j.aiia.2025.07.001
Ya Zhao , Wen Zhang , Liangxiao Zhang , Xiaoqian Tang , Du Wang , Qi Zhang , Peiwu Li
Nodule formation and their involvement in biological nitrogen fixation are critical features of leguminous plants, with phenotypic characteristics closely linked to plant growth and nitrogen fixation efficiency. However, the phenotypic analysis of root nodules remains technically challenging due to their small size, weak texture, dense clustering, and occlusion. To address these challenges, this study constructed a scanner-based imaging platform and optimized data acquisition conditions for high-resolution, high-consistency root nodule images under field conditions. In addition, A hybrid small-object detection method, SCO-YOLOv8s, was proposed, integrating Swin Transformer and CBAM attention mechanisms into the YOLOv8s framework to enhance global and local feature representation. Furthermore, an Otsu segmentation-based post-processing module was incorporated to validate and refine detection results based on geometric features, boundary sharpness, and image entropy, effectively reducing false positives and enhancing robustness in complex scenes. Using this integrated approach, over 3375 nodules were identified from a single plant sample in under 1 min, with extracted phenotypic features such as diameter, color, and texture. A total of 10,879 high-quality annotated images were collected from 39 peanut varieties across 14 provinces and 31 soybean varieties across 12 provinces in China, addressing the current lack of large-scale datasets for legume root nodules. The SCO-YOLOv8s model achieved a precision of 97.29 %, a mAP of 98.23 %, and an overall identification accuracy of 95.83 %. This integrated approach provides a practical and scalable solution for high-throughput nodule phenotyping, and may contribute to a deeper understanding of nitrogen fixation mechanisms.
{"title":"Development of an enhanced hybrid attention YOLOv8s small object detection method for phenotypic analysis of root nodules","authors":"Ya Zhao , Wen Zhang , Liangxiao Zhang , Xiaoqian Tang , Du Wang , Qi Zhang , Peiwu Li","doi":"10.1016/j.aiia.2025.07.001","DOIUrl":"10.1016/j.aiia.2025.07.001","url":null,"abstract":"<div><div>Nodule formation and their involvement in biological nitrogen fixation are critical features of leguminous plants, with phenotypic characteristics closely linked to plant growth and nitrogen fixation efficiency. However, the phenotypic analysis of root nodules remains technically challenging due to their small size, weak texture, dense clustering, and occlusion. To address these challenges, this study constructed a scanner-based imaging platform and optimized data acquisition conditions for high-resolution, high-consistency root nodule images under field conditions. In addition, A hybrid small-object detection method, SCO-YOLOv8s, was proposed, integrating Swin Transformer and CBAM attention mechanisms into the YOLOv8s framework to enhance global and local feature representation. Furthermore, an Otsu segmentation-based post-processing module was incorporated to validate and refine detection results based on geometric features, boundary sharpness, and image entropy, effectively reducing false positives and enhancing robustness in complex scenes. Using this integrated approach, over 3375 nodules were identified from a single plant sample in under 1 min, with extracted phenotypic features such as diameter, color, and texture. A total of 10,879 high-quality annotated images were collected from 39 peanut varieties across 14 provinces and 31 soybean varieties across 12 provinces in China, addressing the current lack of large-scale datasets for legume root nodules. The SCO-YOLOv8s model achieved a precision of 97.29 %, a mAP of 98.23 %, and an overall identification accuracy of 95.83 %. This integrated approach provides a practical and scalable solution for high-throughput nodule phenotyping, and may contribute to a deeper understanding of nitrogen fixation mechanisms.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 1","pages":"Pages 12-43"},"PeriodicalIF":12.4,"publicationDate":"2025-07-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144722714","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-07DOI: 10.1016/j.aiia.2025.06.008
Kaixuan Cuan , Feiyue Hu , Xiaoshuai Wang , Xiaojie Yan , Yanchao Wang , Kaiying Wang
Rapid and accurate measurement of body temperature is essential for early disease detection, as it is a key indicator of piglet health. Infrared thermography (IRT) is a widely used, convenient, non-intrusive, and efficient non-contact temperature measurement technology. However, the activities and clustering of group-housed piglets make it challenging to measure the individual body temperature using IRT. This study proposes a method for detecting body temperature in group-housed piglets using infrared-visible image fusion. The infrared and visible images were automatically captured by cameras mounted on a robot. An improved YOLOv8-PT model was proposed to detect both piglets and their key body regions (ears, abdomen and hip) in visible images. Subsequently, the Oriented FAST and Rotated BRIEF (ORB) image registration method and the U2Fusion image fusion network were employed to extract temperatures from the detected body parts. Finally, a core body temperature (CBT) estimation model was developed, with actual rectal temperature serving as the gold standard. The temperatures of three body parts detected by infrared thermography were used to estimate CBT, and the maximum estimated temperature based on these body parts (EBT-Max) was selected as the final result. In the experiment, the YOLOv8-PT model achieved a [email protected] of 93.6 %, precision of 93.3 %, recall of 88.9 %, and F1 score of 91.05 %. The average detection time per image was 4.3 ms, enabling real-time detection. Additionally, the mean absolute errors (MAE) and correlation coefficient between EBT-Max and actual rectal temperature is 0.40 °C and 0.6939, respectively. Therefore, this method provides a feasible and efficient approach for group-housed piglets body temperature detection and offers a reference for the development of automated pig health monitoring systems.
{"title":"Automatic body temperature detection of group-housed piglets based on infrared and visible image fusion","authors":"Kaixuan Cuan , Feiyue Hu , Xiaoshuai Wang , Xiaojie Yan , Yanchao Wang , Kaiying Wang","doi":"10.1016/j.aiia.2025.06.008","DOIUrl":"10.1016/j.aiia.2025.06.008","url":null,"abstract":"<div><div>Rapid and accurate measurement of body temperature is essential for early disease detection, as it is a key indicator of piglet health. Infrared thermography (IRT) is a widely used, convenient, non-intrusive, and efficient non-contact temperature measurement technology. However, the activities and clustering of group-housed piglets make it challenging to measure the individual body temperature using IRT. This study proposes a method for detecting body temperature in group-housed piglets using infrared-visible image fusion. The infrared and visible images were automatically captured by cameras mounted on a robot. An improved YOLOv8-PT model was proposed to detect both piglets and their key body regions (ears, abdomen and hip) in visible images. Subsequently, the Oriented FAST and Rotated BRIEF (ORB) image registration method and the U2Fusion image fusion network were employed to extract temperatures from the detected body parts. Finally, a core body temperature (CBT) estimation model was developed, with actual rectal temperature serving as the gold standard. The temperatures of three body parts detected by infrared thermography were used to estimate CBT, and the maximum estimated temperature based on these body parts (EBT-Max) was selected as the final result. In the experiment, the YOLOv8-PT model achieved a [email protected] of 93.6 %, precision of 93.3 %, recall of 88.9 %, and F1 score of 91.05 %. The average detection time per image was 4.3 ms, enabling real-time detection. Additionally, the mean absolute errors (MAE) and correlation coefficient between EBT-Max and actual rectal temperature is 0.40 °C and 0.6939, respectively. Therefore, this method provides a feasible and efficient approach for group-housed piglets body temperature detection and offers a reference for the development of automated pig health monitoring systems.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"16 1","pages":"Pages 1-11"},"PeriodicalIF":8.2,"publicationDate":"2025-07-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144595478","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-24DOI: 10.1016/j.aiia.2025.06.007
Xiangyu Zhao , Fuzhen Sun , Jinlong Li , Dongfeng Zhang , Qiusi Zhang , Zhongqiang Liu , Changwei Tan , Hongxiang Ma , Kaiyi Wang
Plant breeding stands as a cornerstone for agricultural productivity and the safeguarding of food security. The advent of Genomic Selection heralds a new epoch in breeding, characterized by its capacity to harness whole-genome variation for genomic prediction. This approach transcends the need for prior knowledge of genes associated with specific traits. Nonetheless, the vast dimensionality of genomic data juxtaposed with the relatively limited number of phenotypic samples often leads to the “curse of dimensionality”, where traditional statistical, machine learning, and deep learning methods are prone to overfitting and suboptimal predictive performance. To surmount this challenge, we introduce a unified Variational auto-encoder based Multi-task Genomic Prediction model (VMGP) that integrates self-supervised genomic compression and reconstruction with multiple prediction tasks. This approach provides a robust solution, offering a formidable predictive framework that has been rigorously validated across public datasets for wheat, rice, and maize. Our model demonstrates exceptional capabilities in multi-phenotype and multi-environment genomic prediction, successfully navigating the complexities of cross-population genomic selection and underscoring its unique strengths and utility. Furthermore, by integrating VMGP with model interpretability, we can effectively triage relevant single nucleotide polymorphisms, thereby enhancing prediction performance and proposing potential cost-effective genotyping solutions. The VMGP framework, with its simplicity, stable predictive prowess, and open-source code, is exceptionally well-suited for broad dissemination within plant breeding programs. It is particularly advantageous for breeders who prioritize phenotype prediction yet may not possess extensive knowledge in deep learning or proficiency in parameter tuning.
{"title":"VMGP: A unified variational auto-encoder based multi-task model for multi-phenotype, multi-environment, and cross-population genomic selection in plants","authors":"Xiangyu Zhao , Fuzhen Sun , Jinlong Li , Dongfeng Zhang , Qiusi Zhang , Zhongqiang Liu , Changwei Tan , Hongxiang Ma , Kaiyi Wang","doi":"10.1016/j.aiia.2025.06.007","DOIUrl":"10.1016/j.aiia.2025.06.007","url":null,"abstract":"<div><div>Plant breeding stands as a cornerstone for agricultural productivity and the safeguarding of food security. The advent of Genomic Selection heralds a new epoch in breeding, characterized by its capacity to harness whole-genome variation for genomic prediction. This approach transcends the need for prior knowledge of genes associated with specific traits. Nonetheless, the vast dimensionality of genomic data juxtaposed with the relatively limited number of phenotypic samples often leads to the “curse of dimensionality”, where traditional statistical, machine learning, and deep learning methods are prone to overfitting and suboptimal predictive performance. To surmount this challenge, we introduce a unified Variational auto-encoder based Multi-task Genomic Prediction model (VMGP) that integrates self-supervised genomic compression and reconstruction with multiple prediction tasks. This approach provides a robust solution, offering a formidable predictive framework that has been rigorously validated across public datasets for wheat, rice, and maize. Our model demonstrates exceptional capabilities in multi-phenotype and multi-environment genomic prediction, successfully navigating the complexities of cross-population genomic selection and underscoring its unique strengths and utility. Furthermore, by integrating VMGP with model interpretability, we can effectively triage relevant single nucleotide polymorphisms, thereby enhancing prediction performance and proposing potential cost-effective genotyping solutions. The VMGP framework, with its simplicity, stable predictive prowess, and open-source code, is exceptionally well-suited for broad dissemination within plant breeding programs. It is particularly advantageous for breeders who prioritize phenotype prediction yet may not possess extensive knowledge in deep learning or proficiency in parameter tuning.</div></div>","PeriodicalId":52814,"journal":{"name":"Artificial Intelligence in Agriculture","volume":"15 4","pages":"Pages 829-842"},"PeriodicalIF":8.2,"publicationDate":"2025-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144534365","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号